October 31, 2019
Machine learning in retail: upgrading your store
As technology evolves, it gives the possibility for the retail industry to reinvent itself and find a more optimized way of functioning. There is plenty of consumer data available for grabs. Aggregating this data and feeding it into machine learning algorithms gives valuable insights into consumer preferences. This allows personalizing products and the way they are sold. However, there are other aspects where data analytics helps retail brands remain competitive.
Including machine learning consulting into your practice allows optimizing delivery routes, which reduces cost and ecological footprint. It helps establish more effective inventory management, reduce waste, and prevent out-of-stock situations. Machine learning also helps find the optimal prices for your products and change them dynamically to remain relevant.
Nowadays, customers have better access to information and platforms to express their dissatisfaction with the retailers who disappoint. Simultaneously, retailers have more opportunities to tailor their business to consumers’ needs and gain a competitive advantage.
Targeting ads are those based on consumer behavior, such as buying or clicking on particular products. There are specific advertising applications that can help in this matter.
For example, Match2One is an application that uses machine learning to identify and target individuals who are likely to purchase from a particular retailer. When deployed on an ecommerce site, it analyzes customer behavior and generates targeted ads accordingly. Match2One can also help with re-targeting: if the user checked an item but did not convert, the application will keep reminding them of this item by advertising it.
This allows changing website content dynamically for different consumers based on such variables as location, demographics, and purchase history.
Bookmark’s AI Design Assistant called Aida helps retailers build a website by predicting which elements, colors, and other design parts it should feature to resonate with the audience most. When the website is up and running and as users get to visiting it, Aida gathers more context and offers recommendations to improve the website.
It suggests items that consumers are very likely to need, even if they have never searched for them in the first place.
Amazon is famous for its recommendation engine, which is responsible for 55% of Amazon’s sales. Its machine learning algorithm considers not only consumer behavior such as placed orders, clicked and searched items, but also similar customers’ purchasing behavior. For example, if you are searching for a present for Mother’s Day, the recommendation algorithm will consider your purchasing behavior in addition to the behavior of other people with similar attributes searching for the same type of present.
Price optimization techniques help retailers price their products in a way that ensures profit without overpricing to the degree where consumers are not willing to pay. There is no single pricing formula that works for every retailer in every industry.
Machine learning gives the possibility of dynamically adjusting the price based on changing circumstances, as it constantly gathers new context and puts it in action. If you employ machine learning, the algorithm will constantly crawl the web searching for prices of similar products offered by your competitors, their hot deals, and the overall pricing history of particular products.
Machine learning helps decide on promotions. It can analyze various factors including product type, demand, and the size of discount. It can even incorporate factors that cannot be processed by traditional price optimization algorithms, such as shelf placement of the product and marketing campaigns in place. Machine learning suggests promotions that are likely to yield a satisfactory ROI as well as those that are risky.
Airbnb is using machine learning to recommend prices for potential hosts, considering a multitude of variables such as photos and locations with weights assigned to them. The algorithm is also able to predict the likelihood of a particular property being booked based on real-time market data.
Efficient route planning requires analyzing real-time data and adjusting on the fly as new information comes in. Machine learning coupled with the internet of things can do just that by analyzing data coming from different sensors. This includes data such as weather conditions, traffic conditions, and the location of each customer.
Optimizing delivery routes brings the following benefits:
Anheuser-Busch adopted machine learning for route optimization in a pilot program covering two US cities. The company used machine learning algorithms for daily deliveries. The algorithms considered factors such as weather conditions and the driver’s experience to suggest the best delivery time for each consumer.
After a few months of running the pilot program, Anheuser-Busch reviewed the success metrics including driver satisfaction and working hours. As a result, the company decided to extend this navigation approach to all of its US wholesalers.
The delivery process is bound by many constraints. Some of them are related to human nature. For example, there are only that many hours that a human driver can drive without interruption. Having two drivers in one truck can be expensive, and waiting for one driver to rejuvenate can significantly prolong the trip.
Different retailers are already experimenting with self-driving vehicles for their last-mile deliveries. Amazon is testing Amazon Scout, a self-driving electric device the size of a cooler. Amazon started with six such devices delivering packages on weekdays in daylight. Initially, human employees accompanied the scouts to ensure they safely navigated around pets and pedestrians.
Kroger is partnering with technology startup Nuro to deliver groceries in self-driving vehicles seven days a week. After placing an order, the customer can choose same-day or next-day delivery.
Excessive stock results in a waste of money as well as headaches as to how to get rid of it through promotions. At the same time, stock shortages damage brand reputation and sway consumers to competitors’ side. Fashion house Burberry sparked outrage after burning £28 million worth of its overstocked items.
Retailers are expected to provide the right product at the right time in the appropriate condition based on consumers’ demand. This requires in-depth supply chain and delivery analyses on the one hand, and the study of consumer buying tendencies on the other. Given the current state of affairs, this is more of a chance than a proper inventory management strategy. However, machine learning opens the possibility to apply big data in retail, so that retailers can develop effective inventory management strategies.
Aston Martin used machine learning to analyze huge historical datasets gathered by their spare parts department over decades. The algorithm picked up eight completely new behavior patterns, which helped to adjust the stocking of spare parts and eventually reduce inventory costs by 18%.
This is an inventory management system solely based on AI and machine learning. It analyzes real-time data coming from a multitude of sources. It differs from traditional systems in that it can predict different scenarios and suggest as well as actually perform actions. This system can be configured to either work autonomously on some decisions or to make suggestions that human employees will have to review and approve.
To this day, most cameras at retail stores are mainly used for security reasons, and they require an employee to watch the monitor and take actions when needed. Those cameras do not add value without human intervention.
Machine vision offers the retail industry an opportunity to use cameras to gather data such as consumer demographics, suspicious behavior, etc. and store it with minimal human intervention. These cameras need to be set up and integrated into the rest of an outlet’s AI initiatives, but the opportunities it opens are worth the trouble. These cameras will be able to:
If suspicious behavior is detected, an AI system can alert store managers or the security team to what is going on through sending the respective video. Human employees will review the footage and determine the best course of action in this situation.
Facial recognition software scans “faceprints” in a similar way to fingerprints, enabling machine learning algorithms to detect known shoplifters the moment they enter the store.
Facewatch is a UK facial recognition technology company that has built its database of potential thieves. Facewatch maintains the ownership of the database, but retailers can enrich it with their own ‘subjects of interest’. Recently, the company has been testing its facial recognition software at grocery stores. The CEO of Facewatch claims that grocery retailers were able to reduce theft up to 40-50% during the 12-week trial period.
Computer vision technology can be used to spot particular actions, for example, detecting if someone picked an item and hid it in their pocket or backpack. Additionally, cameras will be able to detect any other suspicious behavioral patterns which are common among thieves.
Those cameras will also discourage theft among employees. A camera capturing the checkout assistant will detect any “sweethearting” when the assistant avoids scanning some items as a favor to their friend or a frequent customer. The assistant can achieve this by stacking items on top of each other, covering the barcode, or even by skipping the scanning entirely.
Founder and CEO of Everseen, a retail technology firm
If retailers want to reliably detect theft with computer vision, it will take more than high-resolution cameras and algorithms. What you also need is a huge number of labeled examples of visual data representing different patterns of theft behavior.
A Japanese technology startup, Vaak, has recently developed AI software able to catch shoplifters in action and alert store managers. The CEO of Vaak admitted to training machine learning algorithms on 100,000 hours of video data depicting over 100 behavioral aspects including hand movements, facial expressions, walking patterns, etc.
AI and machine learning are swiftly invading the retail space, opening new opportunities for retailers to differentiate themselves from their competitors. However, to fully benefit from machine learning, there are adoption barriers to consider.
You will need a way to gather and aggregate real-time data about your customers, products, competition, and relevant external events. Much of this data can be obtained through web crawling. Bigger datasets yield better results, but even a smaller set is a reasonable way to start. Additionally, it is preferable to install machine vision-based cameras as they help to gather data from your store.
If all this seems like unnecessary hustle, consider the benefits machine learning brings:
Learn how Itransition delivered retail BI and deployed an ML-based customer analytics solution now processing 10TB of data.
Look up the various use cases, challenges, and benefits of applying IoT in retail in Itransition’s guide.
Beacons in retail are important components of proximity marketing and usability strategies. Learn how to use them to increase ROI and avoid risks.
How can business intelligence turn data into higher profits for retailers? Read on to find our BI consultant’s answer to this.
Learn why modern retail needs big data ecommerce solutions to thrive in the age of personalized omnichannel customer experience.
We tap into big data and its business impacts to show how to keep it under control and avoid common big data management pitfalls.